July 1997 ~ Volume 5, Number 4
Trust Introduces New Translation Criteria
As the integration of patient-based outcome measures into all sectors of health care expands, the need arises for instruments capable of capturing data across cultures. In recent years a rapid increase in the number of available translations of both generic and condition-specific instruments has occurred throughout the world.
In cases in which the two languages are similar translation involves only minor modifications to an original version. But in most cases, the process is much more comprehensive and involves the complete translation of the instrument from one language to another. Due to the increase in demand for such instruments, both domestically and internationally, this issue of the Bulletin is dedicated to exploring issues related to the adaptation of cross-cultural instruments through translation. This process is indeed complex, and many factors must be considered in beginning such a project. These factors will be addressed in this issue.
The rise in demand for translated instruments is driven in part from the need to aggregate data from two or more cultures in clinical trials of the efficacy and effectiveness of procedures and medications, and from the shift to an increasingly "global" orientation of health care. According to Donald Patrick, PhD, MSPH, Professor, University of Washington, and member of the Trust's Scientific Advisory Committee (SAC), "these global perspectives require outcome measures, including self-reported health and quality of life, that can be used to support claims, evaluate treatments, and monitor trends." There are at least two reasons for the increase in the search for evidence globally. One is the growing demand for evidence of effectiveness and efficacy as soon as possible by patients and purchasers (including governments) before procedures or pharmaceuticals will be approved for provision and/or payments. Treatments such as those for breast cancer and Parkinson's disease have come to depend on studies done in other countries to accumulate sufficient evidence for consideration. The second reason is that for some conditions only international studies will provide researchers with enough of the cases required to be confident of the results generated by these studies, as would be the case with certain rare disorders.
Because different cultures adopt different values and norms, often the mere linguistic translation of an instrument is not sufficient- an instrument must also be appropriately adapted for the culture in question. To no surprise, this process is lengthy and costly. Several instruments that were adapted both linguistically and culturally have been submitted to the Trust for possible inclusion in it's portfolio of instruments, but some of these instruments have not been fully documented in terms of their measurement properties.
In light of the above experience, SAC developed in April 1997 a set of criteria related to (a) the process of translation and adaptation, and to (b) the documentation required for translated versions to be considered by the Trust for approval and distribution. SAC will apply these criteria to evaluate instruments submitted to the Trust for approval of the translation only. For SAC to conduct a review of the translated version, the original (source) instrument must have been previously approved by the Trust. The Trust expects that this new process will further enlarge the Trust's portfolio of instruments and increase the availability and accessibility of translated instruments.
SAC believes that Trust members and potential users should be informed of the availability of translated and adapted instruments even though their measurement properties have not been comprehensively evaluated. Such a step is predicted on two provisions: that such instruments meet, at minimum, accepted standards of translation and that the original instrument has been previously approved by the Trust.
The following are the
1. The source instrument and its measurement properties must have been reviewed by the Scientific Advisory Committee and approved by the Medical Outcomes Trust.
2. Evidence has been provided of a minimum of two independent forward translations of the instrument.
3. Evidence has been provided of a minimum of one back translation of the instrument.
4. A detailed description has been provided of the translation process, including a description of the population involved in the evaluation of the appropriateness of the instrument and, whenever possible, a description of the qualifications of the translators used to conduct the translations.
5. The final forward and backward translation document and documentation of modifications made have been provided.
The Trust will identify instruments reviewed and approved using the above process as having met the "Minimal Translation Criteria." Included in packets for such instruments will be the same information that is currently provided in the Trust's existing standard packet: a copy of the questionnaire, scoring and interpretation guidelines, a bibliography, and reprints of published articles pertinent to its development and use.
Because the new Minimal Translation Criteria may raise questions among those interested in translating an instrument, the article on page two of this Bulletin is dedicated to the process of instrument translation; it explains the steps involved in translation and touches on issues to consider when translating.
Medical Outcomes Trust
Approaches to Instrument Translation: Issues to Consider
Researchers in the field of quality of life assessment have demonstrated three principle approaches to instrument adaptation and translation. Of these approaches, the one most commonly used is known as the sequential model. This model, employed with instruments such as the Nottingham Health Profile, the Sickness Impact Profile, the SF-36 Health Survey, and the Quality of Well-Being Scale, involves an established set of procedures which include a forward/back translation, the use of focus groups, and other validation efforts of an original source instrument into the intended language. The EORTC Quality of Life Study Group and the World Organization of General Practitioners and Family Physicians uses a second approach known as the parallel model. This model seeks to establish a basis for cross-national comparability in the original development and validation stages of an instrument. A third model, known as the simultaneous approach, has been employed by the WHO Quality of Life Group for the WHOQOL instrument. After reaching consensus on the domains to be included in the instrument, each country may develop its own questionnaire items using the group's standard set of procedures.
According to researchers at the MAPI Institute in Lyon, France, an organization responsible for the translation of over 70 instruments into 30 languages, the object of translation is to remove the difficulties barring the optimal transfer of the informational, emotional, and stylistic content of the original message. The general process of translating an instrument consists of the following four steps:
- Forward step
- Quality control
- International harmonization
The forward step of translating from source (original) language to target language is initiated by contacting the source author to secure authorization and create an understanding of the underlying concepts for items and dimensions. Once this contact has been made, it should be maintained, as the author is capable of lending valuable insight which could enhance the translation process. Next, the production of several forward translations of the instrument, using translators residing in the target country, familiar with tenets of the field of health outcomes, is recommended. Once these versions are available, a consensus meeting comprised of experts should convene to discuss and evaluate the first translation efforts.
The quality control step exists to ensure that the target version is equivalent to the source version both conceptually and linguistically. This type of control can be achieved through quality ratings or back translations. Quality ratings are based upon conceptual equivalence with the original, clarity, and use of a familiar or colloquial register. Each of these variables is rated on a three-point scale by at least two reviewers, and those variables evaluated as unsatisfactory should then be reevaluated. A back translation is the retranslation of the target language version into the source language version. The objective of back translation is to detect errors of meaning and concept nonequivalence.
The pretest of a translated instrument can occur with either a monolingual lay panel or a bilingual lay panel, but in either case, the objectives remain the same: to measure comprehensibility; to test translation alternatives; to highlight unexpected or undetected errors; and to reveal inappropriate items. Regardless of what panel is chosen, it must consist of heterogeneous individuals. In the monolingual panel, generally the target version is tested through face-to-face interviews or with focus groups. The bilingual panel actually completes both the source and the target versions and items receiving discrepant responses are investigated.
In addition to descriptions of the pretesting populations and qualifications of translators, those at the MAPI Institute suggest that validation reports are kept as well. These reports should note among other things, which items were difficult to translate, and therefore, may need further cultural modification.
The end of the translation process is typically marked by the pretesting of an instrument, however, when a instrument is being translated into several languages at once, an international harmonization meeting comprised of as many bilingual (English) professional translators as possible (for each of the target versions) will ensure that target versions truly are conceptually equivalent.
While the process described above has its strengths - it is intuitively reasonable, flattering, and simple - it is not without its limitations: (1) questionnaires are likely to have weaknesses even in their original version; (2) some items just do not translate well; (3) some items just do not make sense in certain cultural contexts; (4) items important to the population for whom the original version was created may be of little consequence for the target population. Each limitation may require a different solution in order to be overcome, and such solutions, though not always well developed, should not be overlooked or diminished.
In any approach to cross-cultural adaptation or translation, several important considerations should be taken into account, such as the cross-cultural content, the meaning of different health and quality of life domains, the translation of measures from one language to another, and cross-cultural validation. It is also important to determine the purpose of the measurement which may include discrimination between health status of different population groups, evaluation of change over time, or predicting future health status. For additional reading on instrument translation, the following list of publications (used to prepare this article) may serve as a useful reference.
Quality of Life and Pharmacoeconomics in Clinical Trials, Second Edition, edited by B. Spilker. Lippincott-Raven Publishers, Philadelphia 1996
Quality of Life Assessment: International Perspectives, edited by J. Orley and W. Kuyken. Springer-Verlag Publishers, Berlin 1994
The International Assessment of Health-Related Quality of Life: Theory, Translation, Measurement and Analysis, edited by SD. Shumaker and R. Berzon. Rapid Communications, Oxford 1995
Crosscultural Efforts in Health Outcomes Assessment
To illustrate examples of efforts in establishing crosscultural measures, this article of the Bulletin will summarize activities for six well known instruments that measure health outcomes.
Dartmouth COOP Function Charts/WONCA: Developed by a network of clinicians, the COOP Charts evaluate overall patient functioning through the use of pictorial representations of functioning levels. The COOP Charts have been used internationally in the Netherlands, Canada, and Japan, and in 1988, the World Organization of General Practitioners/Family Physicians (WONCA) selected the instrument as the basis for the development of an international system for the measurement of functional status. COOP/WONCA charts are available in Danish, Dutch, Finnish, German, Hebrew, Japanese, Norwegian, Spanish and Urdu. Currently, little research exists pertaining to acceptance, relevance, and meaning of the verbal response levels and symbols used in the WONCA system, however, testing is on-going, and thus far, the revised COOP Charts are reported to detect moderate effects in physical and emotional functioning due to medical illness or health status.
The EuroQol instrument is a multidimensional health related quality of life (HRQL) profile capable of creating a general cardinal index of health. The EuroQol instrument has been translated and evaluated through the parallel approach- it was developed by the EuroQol Group, a multi-country, multi-center, and multi-disciplinary organization which convened in 1987 with the objective of developing a standardized instrument to describe quality of life. This instrument is currently available in English, Spanish, Catalian, Dutch, Finnish, Norwegian, Swedish, French, and German, and is intended for use in concert with other HRQL measures. The instrument has been available since 1990 and currently, there is preliminary support for the comparability of health valuations across the countries considered. However, there is little published regarding methods of adaptation, and similarly, psychometric properties await further evaluation.
Nottingham Health Profile (NHP): A self-administered measure assessing the effects of illness on behavior, the NHP was originally developed in England and has been adapted, through the sequential approach, into Italian, French, Dutch, Swedish, and Spanish. Psychometric testing has been documented for all but the Italian version. In terms of translation, the process was systematic and complete, including forward and backward translation and an iterative review of the quality of the linguistic and semantic equivalence. As for scale equivalence of adapted versions, this appears to be preliminary due to a lack of coordination in study design and psychometric analyses across countries. Metric equivalence is provided from the generally high reproducibility of item ranks. Operational equivalence of language-adapted versions have yet to be reported comparing telephone, self-administered, and postal data collection methods.
SF-36 Health Survey: A generic measure which evaluates eight health domains in a 36-item questionnaire, the SF-36 is a short form developed in the Medical Outcomes Study. In 1991, the International Quality of Life Assessment (IQOLA) Project was launched to translate, validate, and norm the SF-36 for further use in multinational clinical trials and other international studies. Based at The Health Institute, New England Medical Center, the project originally included sponsored investigators from 14 countries: Australia, Belgium, Canada, Denmark, France, Germany, Italy, Japan, the Netherlands, Norway, Spain, Sweden, the UK and the US (Spanish). Researchers from more than 25 other countries are also translating and validating the SF-36 following IQOLA methods. Four major stages of activity are included: translation following a standard protocol; formal psychometric tests of scaling and scoring assumptions; validation and norming studies to evaluate the equivalence of interpretations across countries; and documentation of scoring algorithms and interpretation guidelines. Trust approved language versions of the SF-36 include translations for Germany Italy, Spain and Sweden, and English-language adaptations for Australia/New Zealand, Canada, and the United Kingdom.
Sickness Impact Profile (SIP): A measure of perceived health status, the SIP assesses the impact of sickness on daily activities. The SIP was developed in the United States, and translated versions include Anglican, Swedish, German, French, Danish, Norwegian, and Dutch. Psychometric evaluations of the Anglican, Dutch, and Swedish versions have been extensive. Item rankings between the original version and international versions suggest a high degree of conceptual agreement. The performance of many of these versions in separating chronic versus acute health conditions, and levels of illness severity indicate that the adoptions have been largely successful. Additionally, psychometric evaluation documented in research for the Dutch and Swedish versions is quite advanced. However, overall, both operational and metric equivalence in language versions have yet to be extensively documented, and this is due to the incomplete formal testing of cross-cultural equivalence (currently ongoing). The original SIP instrument (U.S. English) has been approved by the Trust.
The simultaneous approach to instrument development is exemplified by the World Health Organization (WHO), which, in collaboration with 15 international centers, developed the WHO Quality of Life Assessment (WHOQOL). The WHOQOL-100 and WHOQOL-BREF incorporate many dimensions of life and can determine impact of disease and health intervention on quality of life. Because this instrument is based on cultural norms of various groups, as opposed to the traditional method of basing item selection on one set of culturally homogeneous data, it truly is a cross-cultural measure. Furthermore, it is capable of capturing quality of life data in countries which lack validated measures. While testing of the psychometric properties of the WHOQOL instruments is ongoing, this collaborative effort is a fine example of what is possible when we start to think beyond domestic concepts of quality of life and well-being. But until resources and vision allow for the modeling of such a collaborative spirit and global philosophy, individuals interested in cross-cultural measures of quality of life and health will have to rely on translation and adaptation efforts.
Patrick DL, Wild DJ, Johnson ES, et al. Cross-cultural validation of quality of life measures. Orley J, Kuyken, W, eds. Quality of Life Assessments: International Perspectives. Berlin: Springer-Verlag, 1994.
Anderson RT, Aaronson NK, Leplege AP, et al. International use and application of generic health related quality of life instruments. Spilker B, ed. Quality of Life and Pharmacoeconomics in Clinical Trials, Second Edition. Philadelphia: Lippincott-Raven, 1996
Anderson RT, Aaronson NK, Wilkin, D. Critical review of the international assessments of health-related quality of life: generic instruments. Shumaker SA, Berzon RA, eds. The International Assessment of Health-related Quality of Life: Theory, Translation, Measurement, and Analysis. Oxford: Rapid Communications, 1995.
Patrick DL (1997). Crosscultural assessment of health and quality of life outcomes. In press.
© 1997 MEDICAL OUTCOMES TRUST